Indian Sign Language Recognition: A Comparative Analysis Using CNN and RNN Models

The deaf and hard-of-hearing people in India uses a distinctive form of communication called Indian Sign Language (ISL). Effective communication and inclusion require the ability to recognise and comprehend ISL gestures. Deep learning algorithms have advanced recently, opening the door for autonomous systems that can recognise sign language. In order to recognise Indian Sign Language, this research compares and contrasts Convolutional Neural Networks (CNN) and Recurrent Neural Networks (RNN), two well-known deep learning models. The proposed study focuses on developing an accurate and efficient system capable of recognizing and interpreting ISL gestures. The experimental findings show that both CNN and RNN models are equally effective and reliable for recognising Indian Sign Language. The RNN model excelled in capturing the temporal dynamics and sequential patterns of ISL movements, while the CNN model achieved excellent accuracy in capturing the spatial aspects of ISL gestures. A comparison study was also done to assess how well these models performed in terms of accuracy, precision, recall, and F1-score. In this work, CNN and RNN models are compared for the recognition of Indian Sign Language. The findings emphasise the need of taking both geographical and temporal information into account when creating precise and effective sign language recognition systems. The results offer insightful information for academics and professionals striving to improve accessibility and communication for the deaf and hard-of-hearing people in India. In order to further enhance the performance of ISL recognition systems, future research paths may investigate hybrid CNN-RNN architectures and incorporate other modalities like depth information or hand motion tracking.